datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
NuminaMath-1.5-RL-Verifiable
Dataset Card for NuminaMath-1.5-RL-Verifiable
Dataset Summary
NuminaMath-1.5-RL-Verifiable is a curated subset of the NuminaMath-1.5 dataset, specifically filtered to support reinforcement learning applications requiring verifiable outcomes. This collection consists of 131,063 math word problems from the original dataset that meet strict filtering criteria: all problems have definitive numerical answers, validated problem statements and solutions, and come from… See the full description on the dataset page: https://huggingface.co/datasets/nlile/NuminaMath-1.5-RL-Verifiable.NuminaMath-20k
Merged Math Datasets (100k Subset)
This dataset combines multiple mathematical datasets for training and evaluation purposes. This version contains a shuffled 100k subset of the training data for faster experimentation.
Dataset Description
A comprehensive collection of mathematical problems and solutions from various sources, organized into training and multiple test subsets.
Dataset Structure
Training Set
Size: 20000 examples
Fields: source… See the full description on the dataset page: https://huggingface.co/datasets/weijiezz/NuminaMath-20k.NuminaMath-full
Merged Math Datasets (Full)
This dataset combines multiple mathematical datasets for training and evaluation purposes. This version contains the complete training dataset from all source datasets.
Dataset Description
A comprehensive collection of mathematical problems and solutions from various sources, organized into training and multiple test subsets.
Dataset Structure
Training Set
Size: 217389 examples
Fields: source, question, answer
Sources:… See the full description on the dataset page: https://huggingface.co/datasets/weijiezz/NuminaMath-full.NuminaMath-100k
Merged Math Datasets (100k Subset)
This dataset combines multiple mathematical datasets for training and evaluation purposes. This version contains a shuffled 100k subset of the training data for faster experimentation.
Dataset Description
A comprehensive collection of mathematical problems and solutions from various sources, organized into training and multiple test subsets.
Dataset Structure
Training Set
Size: 100000 examples
Fields: source… See the full description on the dataset page: https://huggingface.co/datasets/weijiezz/NuminaMath-100k.NuminaMath-CoT-decontaminated-filteredDecontaminated version of AI-MO/NuminaMath-CoT/train that remove the collided data from math/test, gsm8k/test, olympiadbench/test, minerva_math/test, college_math/test, mmlu_stem/test, gaokao, amc23, aime24 and math500.
Aligned with flatlander1024/QwQ-LongCoT-130K-decontaminated NuminaMath
Total number of rows: 102238
NuminaMath-CoT-cn_k12-20000-old本数据集为NuminaMath-CoT在cn_k12类别下前20000条数据中进行的长CoT文本生成尝试。
里面有两个数据文件，completion.jsonl和distilled.jsonl。
其中，completion.jsonl是gemma2-27b-it生成的数据，因为gemma2-27b-it最后得到的结果可能提取不出\boxed{}里面的内容等因素，所以最后生成了约18.6k的数据。
而distilled.jsonl是基于上面的回答成功的问题，使用DeepSeek-R1-Distill-Qwen-32B生成的数据（其实不应该这样，之后有机会再从20000条里面直接生成）。
因为NuminaMath-CoT的cn_k12数据大多是一些综合应用题，一条数据中会有多个子问题，回答的时候也要分别回答。对问题回答的答案进行验证较难，尚未进行答案验证，需要下载下来之后再进行更精细化的处理。
completion.jsonl中的数据标签描述：
source (String)：数据来源类别，为原始标签
problem (String)：数学问题文本，为原始标签
solution… See the full description on the dataset page: https://huggingface.co/datasets/tttonyyy/NuminaMath-CoT-cn_k12-20000-old.NuminaMath-CoT-filteredDataset that contains problems that appears in both QwQ-LongCoT-130K-cleaned and NuminaMath-CoT. There are approximately 100k problems where the solution is in plain-CoT manner.
NUMINA
NUMINA: A Natural Understanding Benchmark for Multi-dimensional Intelligence and Numerical Analysis
We introduce NUMINA, the first benchmark focusing on multi-dimensional intelligence and fine-grained numerical reasoning in 3D multimodal environments.
Overview
NUMINA dataset is built upon the ScanNet dataset, a large-scale collection of RGB-D reconstructed indoor scenes.
Each NUMINA sample is grounded in ScanNet 3D geometry and object annotations, enabling evaluation… See the full description on the dataset page: https://huggingface.co/datasets/fengshun124/NUMINA.numinamath_verifiable_cleanedAdapt from https://huggingface.co/datasets/AI-MO/NuminaMath-CoT and filtered problems with verifiable answers.
Removed duplicates and decontaminated from test datasets.
Total number of rows: 678759
NuminaMath-longcot-cot-combinedContains Decontaminated version of AI-MO/NuminaMath-CoT/train and the decontaminated version of gghfez/QwQ-LongCoT-130K-cleaned.
Remove duplicates and merged them into 1 data with 2 different solution rows.
Total number of rows: 87057
NuminaMath-1.5-RL-Verifiable
Dataset Card for NuminaMath-1.5-RL-Verifiable
Dataset Summary
NuminaMath-1.5-RL-Verifiable is a curated subset of the NuminaMath-1.5 dataset, specifically filtered to support reinforcement learning applications requiring verifiable outcomes. This collection consists of 131,063 math word problems from the original dataset that meet strict filtering criteria: all problems have definitive numerical answers, validated problem statements and solutions, and come from… See the full description on the dataset page: https://huggingface.co/datasets/gravermistakes/NuminaMath-1.5-RL-Verifiable.